How Many Conversions You Need to Optimise Google Shopping

As a working rule, a Google Shopping or Performance Max campaign needs somewhere around 15 to 30 conversions in a rolling 30 day window before Google's bidding has enough signal to optimise properly, and closer to 50 conversions a month before you can trust a target ROAS to behave predictably. Below roughly 10 conversions a month the algorithm is largely guessing, and your results will swing about week to week for reasons that have nothing to do with your settings.

That is the short answer. The longer answer matters more, because "conversions" is doing a lot of heavy lifting in that sentence. It means conversions recorded in the campaign, in the last 30 days, attributed to the campaign, on actions you have marked as primary. Not conversions across your whole account. Not conversions in Google Analytics you never imported. Not sales that happened on the phone and never made it back into the system. Get that part wrong and you can have plenty of real sales while the campaign is still starved of data.

Why the algorithm needs a minimum amount of data at all

Smart Bidding is a prediction engine. Every time someone is about to see your product, Google predicts how likely that click is to convert and how much that conversion is likely to be worth, then bids accordingly. Those predictions are built from patterns: device, time of day, location, query intent, audience signals, previous site behaviour, the product itself.

To find a pattern you need examples. With five conversions a month the system has five examples to learn from, spread across dozens of products and hundreds of query variations. There is no pattern to find, so it falls back on broader, less specific models and your results look random. With 50 conversions a month it can start to see that certain product types convert better on mobile in the evening, or that a particular price band carries a higher average order value.

This is also why the advice "give it two weeks" exists. The learning period after a significant change is typically about one to two weeks, but it is really a data period rather than a time period. A campaign generating 60 conversions a month gets through learning in days. A campaign generating six will still be effectively learning a month later.

Conversion volume and what you can realistically do

Here is how we think about it when we take on a new Shopping account. The figures are guidance rather than hard switches, and Google does not publish a single official cut off that applies to every account, but these bands match what we see in practice.

Conversions per 30 days	What the campaign can support	Sensible bidding strategy
0 to 5	Very little. Results will be noisy and unrepeatable.	Manual CPC or maximise clicks, with tight bids and a small product set
5 to 15	Basic learning. Enough to start, not enough to trust targets.	Maximise conversion value with no target, or maximise conversions
15 to 30	Reasonable signal. The algorithm can find repeatable patterns.	Maximise conversion value, introduce a gentle target ROAS once stable
30 to 50	Solid. Target ROAS becomes meaningful and adjustable.	Target ROAS, reviewed fortnightly
50+	Strong. You can segment, test and read results with confidence.	Target ROAS, plus considered campaign splits if there is a real reason

Notice the bottom two rows are where structural changes become safe. Most of the damage we see in small business Shopping accounts comes from someone applying a big account structure to a small account: five campaigns, each with 40 quid a week, each generating two conversions a month. Five campaigns learning badly always lose to one campaign learning well.

Campaign level data, not account level

This is the single most common misunderstanding. Business owners tell us they get 80 orders a month, so surely Google has plenty to work with. Then we look in the account and find those 80 orders split across a Search campaign, a branded campaign, three Shopping campaigns and organic traffic. Each Performance Max campaign might be seeing eight or nine conversions.

Smart Bidding does borrow signals from across the account and from Google's wider models, so a campaign in a mature account with a long history is not starting from absolute zero. But the campaign's own recent conversion data is what drives its specific bidding decisions. Treat the campaign as the unit that needs feeding.

Practical consequences:

• Consolidate. If you have three Shopping campaigns and none of them clears 30 conversions a month, merge them.
• Do not split by brand unless each brand can stand on its own data.
• Do not create a separate campaign for every product category because it feels tidy. Tidy is not a performance strategy.
• Split only when you have a genuine business reason (very different margins, a separate budget that must not be raided, different countries) and enough volume to support both halves afterwards.

Asset groups and listing groups do not fragment data

Within a single Performance Max campaign you can use multiple asset groups and listing groups to organise products. This does not split your bidding data in the same damaging way that separate campaigns do, because the campaign still optimises as one unit. So if you want structure without starving the algorithm, do it inside the campaign rather than by creating new ones.

Are your conversions even being counted properly?

Before you conclude that you have too few conversions, check that you are not losing real ones. We see this constantly. The data problem is often a tracking problem wearing a costume.

Fixing tracking is the cheapest way to increase your conversion count, because the sales already happened. It is normally the first thing we look at during a free Google Ads audit, and it is surprising how often a campaign that appeared to be failing was simply not being told when it had succeeded.

What to do when you genuinely do not have enough conversions

Plenty of good small businesses sell 10 or 15 items a month online. That is a real business and Shopping can still work. You just cannot run it the way a high volume retailer would.

1. Start with a strategy that does not need a target

Maximise conversion value without a target ROAS, or maximise conversions without a target CPA, lets the system spend the budget and gather data rather than throttling itself trying to hit a number it cannot yet predict. Add the target only once you have a stable month of data and a realistic idea of what the account actually returns.

2. Narrow the product set, not the campaign count

If budget is thin, exclude products that will never convert (out of stock lines, one off oddments, very low margin items) so spend concentrates on the products most likely to produce conversions. Concentration builds data. Dispersion destroys it.

3. Use a longer look back when you assess

With low volume, a 7 day view tells you nothing. Judge on 30 days minimum, ideally 60 to 90 days, and compare like with like periods. One bad week at 12 conversions a month is one missing sale.

4. Consider supporting micro conversions carefully

Add to basket or begin checkout can be tracked as secondary actions to give you diagnostic visibility. Be cautious about making them primary and giving them a value, because you can end up with an algorithm that is excellent at generating baskets nobody pays for. If you do it, use it as a temporary bridge while purchase volume builds, and keep the purchase action primary too.

5. Give it a budget that can actually produce conversions

If your average cost per click is 60p and your conversion rate is two per cent, each conversion costs roughly £30 in clicks. To generate 30 conversions a month you would need around £900 of media spend. If your budget is £200, the honest answer is that you will get about six conversions a month and you should plan around that, using a simpler bidding approach and a tighter product selection, rather than expecting Smart Bidding to perform miracles.

How long to wait before judging a change

Every meaningful change (bid strategy, target, budget of more than about 20 per cent, big feed restructure) restarts learning. During that window performance is genuinely unstable and reading it as a result is a mistake.

1. Days 1 to 7: exploration. Cost per conversion often rises. Do not touch anything.
2. Days 8 to 14: stabilising. Patterns start to appear. Still too early to conclude.
3. Days 15 to 30: the first period worth reading, and only if you have accumulated a reasonable number of conversions in it.
4. Day 30 onwards: compare against the previous comparable 30 days, then make one change, not five.

The discipline that separates accounts that improve from accounts that flounder is change frequency. Low volume accounts should be adjusted less often, not more. Every tweak costs you another learning period, and when you only produce 10 conversions a month you cannot afford to spend half of them re-learning.

Moving a target ROAS without wrecking things

When you do have enough data to run a target ROAS, change it gradually. Moves of roughly 10 to 15 per cent at a time, then a fortnight to settle, will get you somewhere. Jumping a target from 300 per cent to 700 per cent because margins are tight usually collapses impressions and delivery, and then you have no data at all.

Set the initial target from your actual recent performance, not from what you wish it was. If the campaign has been returning around 400 per cent with no target, start the target at around 400 per cent and work upward slowly.

Seasonality, noise and small numbers

With low conversion volume, normal random variation looks like a crisis. If you average 12 conversions a month, a month with 8 and a month with 16 are both entirely ordinary outcomes and neither one means your campaign has broken or fixed itself.

The practical habit is to look at trend lines over quarters, and to keep a simple written log of what you changed and when. When someone asks why performance dipped in the second week of March, the log answers the question. Memory does not.

Seasonal businesses have an extra wrinkle. If your sales concentrate into a few months, your campaign will hit useful data volumes in peak season and drop well below them out of season. Plan for that: run a simpler, tighter setup in the quiet months and avoid restructuring right before your busiest period, because you do not want to be in a learning phase when the traffic arrives.

Feed quality still beats bidding at low volume

Here is the thing most people underrate. When you have very little conversion data, the lever with the biggest impact is not the bid strategy. It is the product feed.

• Titles that lead with the words people actually search for, including brand, product type, key attribute and size or variant.
• Product types and Google product categories filled in properly so Google understands what it is matching.
• GTINs and MPNs where they exist, which improves matching and eligibility.
• Images that are clean, correctly cropped and show the product clearly at thumbnail size.
• Price and availability accurate and updated frequently, because disapprovals and mismatches quietly kill impressions.
• Variants handled sensibly, since how you submit them affects what shoppers see. Our piece on whether Google Shopping shows a price range for variants covers that in detail.

A well built feed increases the volume and relevance of the traffic you can win at a given budget, which increases conversions, which feeds the algorithm. It is the only lever that improves both sides of the problem at once.

Edge cases worth thinking about

High value, low volume products

If you sell £4,000 machines and make six sales a month, you will never hit conventional conversion thresholds. Track add to basket, enquiry form and phone call as supporting actions to give the campaign signal, keep the product set tight, expect a long assessment period and accept that a target ROAS will probably never be stable. Many businesses in this position do better with a well structured Search and Shopping combination managed manually, rather than handing everything to automation.

Long consideration cycles

If people research for three weeks before buying, a short conversion window will chronically under report. Lengthen the window, check your attribution model, and be aware that reported performance today reflects clicks from a fortnight ago.

Mixed lead and sale businesses

Retailers who also take trade enquiries need to decide what they are optimising for and value the actions accordingly. Giving an enquiry a sensible estimated value based on your close rate and average order lets the algorithm weigh both fairly, instead of treating a £5 accessory sale and a £900 trade enquiry as equal.

New accounts with no history

A brand new account has no conversion data anywhere. Give it four to six weeks with a simple setup and a budget that can realistically produce sales, resist the urge to restructure in week two, and build the feed properly while you wait. If you are new to all of this, our guide on how Google Ads work for small businesses is a good starting point.

Common mistakes we see

Mistake	Why it hurts	Do this instead
Five campaigns on a small budget	Splits conversion data so nothing learns	Consolidate into one, use listing groups for structure
Setting an ambitious target ROAS on day one	Throttles delivery before any data exists	Run without a target until volume is stable
Changing something every few days	Permanent learning phase	One change, then wait a full assessment period
Pausing after a bad week	Discards the data already paid for	Judge on 30 days minimum
Tracking that misses phone or offline sales	Under reports real conversions	Import offline conversions, use call tracking
Micro conversions set as primary with values	Optimises for baskets, not revenue	Keep purchase primary, use micro actions as secondary

A simple checklist to work through

1. Count conversions per campaign over the last 30 days, not per account.
2. Check which conversion actions are marked as primary and remove duplicates.
3. Confirm Consent Mode and enhanced conversions are configured.
4. Review your conversion window length against your real buying cycle.
5. If any campaign is under 30 conversions a month, consider consolidating.
6. If under 15, drop the target and let the campaign gather data.
7. Work out budget divided by realistic cost per conversion so expectations match maths.
8. Fix the feed properly before blaming the bidding.
9. Log every change with a date, then stop changing things for a fortnight.

None of this requires expensive software. It requires patience and a bit of honesty about the volume your budget can actually produce. If you are still weighing up whether managed PPC is worth it at your spend level, our UK pricing guide to PPC management costs sets out how the numbers usually work, and you can see our own Google Ads pricing on the packages page.